Embed Notice

@realcaseyrollins @stux

looking at polls really means nothing without some meta analysis. Polls differ wildly in their predictions and historic accuracy... So ill try to break down my own process (done non-rigerously) as a data scientist.

1) look through metaanalysis of polls to find polls with long-history of polling across multiple presidents and most accurately predicted the output correctly
2) collect the top 3 polls that have historically correctly predicted the presidential outcome per state
3) using just those polls assume standard deviation on the output per state to identify what polls indicate certainty in prediction and which ones have an error rate where either candidate would fall within the error
4) look at the "swing" states identified and then calculate the statical outcome of either
5) total the electorial college and see if its a slam dunk for one side or likely to be a pretty tight race, the clearly the biases the more confidently i will lean towards that side.